Optimization Of SLAM Algorithm Based On Inertial Measurements Unit And Laser Point Clouds

SLAM(Simultaneous Localization And Mapping)can provide real-time positioning and environmental maps for mobile robots,and is widely used in fields such as autonomous vehicles and logistics automated guided vehicles.Lidar is one of the main methods for SLAM to obtain environmental information.The rotation estimation of pure Lidar SLAM is unstable,and it is difficult to correct point cloud distortion.Inertial Measurement Unit(IMU)can dynamically provide rotation and velocity data,but cannot construct SLAM maps.This paper uses threedimensional lidar combined with an IMU to complete SLAM.The high-precision environmental point cloud is conducive to estimating the trajectory and constructing a dense map,and the high-frequency inertial state can predict the movement.Among the SLAM algorithms that use lidar combined with IMU,there are typically LOAM and LIO-Mapping algorithms,but these do not take into account the situation where the moving carrier starts when the speed is non-zero,such as the carrier restarts after the program is interrupted during operation causing a higher starting speed.If the carrier starts the SLAM program when it is not stationary,the displacement error of the trajectory obtained by the above algorithm is relatively large.In this paper,the following improvements are made on the basis of LIO-Mapping,and a SLAM algorithm based on IMU and laser point cloud:1.Laser SLAM-assisted estimation of initial state method to calculate the initial IMU inertial velocity corresponding to the first laser point cloud frame and re-integrate the inertial pose.2.The sliding window is used to establish the connection between the point cloud to be optimized,the improved inertial pose and the local map,and the z-axis component of the matching residual rotation Jacobian is modified to align with the gravity direction of the global pose.Through the real road driving environment data of the KITTI public data set,experiments are carried out on various road scenes to prove the laser SLAM-assisted estimation of initial state method and SLAM algorithm based on IMU and laser point cloud proposed are feasible.The comprehensive absolute trajectory error is lower than LIO-Mapping,and the accuracy of displacement estimation is improved by at least 10.49%.